Linux-Media Archive on lore.kernel.org
 help / color / Atom feed
* [PATCH v3 1/1] Documentation: media: Document how to write camera sensor drivers
@ 2020-07-30 16:20 Sakari Ailus
  2020-08-03 15:25 ` Helen Koike
  0 siblings, 1 reply; 5+ messages in thread
From: Sakari Ailus @ 2020-07-30 16:20 UTC (permalink / raw)
  To: linux-media; +Cc: laurent.pinchart, hverkuil

While we have had some example drivers, there has been up to date no
formal documentation on how camera sensor drivers should be written; what
are the practices, why, and where they apply.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
---
since v2:

- More verbose explanation on sensor driver's responsibilities.

- Reword the explanation on power state vs. v4l2_ctrl_handler_setup.

 .../driver-api/media/camera-sensor.rst        | 134 ++++++++++++++++++
 Documentation/driver-api/media/csi2.rst       |   2 +
 Documentation/driver-api/media/index.rst      |   1 +
 3 files changed, 137 insertions(+)
 create mode 100644 Documentation/driver-api/media/camera-sensor.rst

diff --git a/Documentation/driver-api/media/camera-sensor.rst b/Documentation/driver-api/media/camera-sensor.rst
new file mode 100644
index 000000000000..2121586e8ede
--- /dev/null
+++ b/Documentation/driver-api/media/camera-sensor.rst
@@ -0,0 +1,134 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+Writing camera sensor drivers
+=============================
+
+CSI-2
+-----
+
+Please see what is written on :ref:`MIPI_CSI_2`.
+
+Handling clocks
+---------------
+
+Camera sensors have an internal clock tree including a PLL and a number of
+divisors. The clock tree is generally configured by the driver based on a few
+input parameters that are specific to the hardware:: the external clock frequency
+and the link frequency. The two parameters generally are obtained from system
+firmware. No other frequencies should be used in any circumstances.
+
+The reason why the clock frequencies are so important is that the clock signals
+come out of the SoC, and in many cases a specific frequency is designed to be
+used in the system. Using another frequency may cause harmful effects
+elsewhere. Therefore only the pre-determined frequencies are configurable by the
+user.
+
+Frame size
+----------
+
+There are two distinct ways to configure the frame size produced by camera
+sensors.
+
+Freely configurable camera sensor drivers
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+Freely configurable camera sensor drivers expose the device's internal
+processing pipeline as one or more sub-devices with different cropping and
+scaling configurations. The output size of the device is the result of a series
+of cropping and scaling operations from the device's pixel array's size.
+
+An example of such a driver is the smiapp driver (see drivers/media/i2c/smiapp).
+
+Register list based drivers
+~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+Register list based drivers generally, instead of able to configure the device
+they control based on user requests, are limited to a number of preset
+configurations that combine a number of different parameters that on hardware
+level are independent. How a driver picks such configuration is based on the
+format set on a source pad at the end of the device's internal pipeline.
+
+Most sensor drivers are implemented this way, see e.g.
+drivers/media/i2c/imx319.c for an example.
+
+Frame interval configuration
+----------------------------
+
+There are two different methods for obtaining possibilities for different frame
+intervals as well as configuring the frame interval. Which one to implement
+depends on the type of the device.
+
+Raw camera sensors
+~~~~~~~~~~~~~~~~~~
+
+Instead of a high level parameter such as frame interval, the frame interval is
+a result of the configuration of a number of camera sensor implementation
+specific parameters. Luckily, these parameters tend to be the same for more or
+less all modern raw camera sensors.
+
+The frame interval is calculated using the following equation::
+
+	frame interval = (analogue crop width + horizontal blanking) *
+			 (analogue crop height + vertical blanking) / pixel rate
+
+The formula is bus independent and is applicable for raw timing parameters on
+large variety of devices beyond camera sensors. Devices that have no analogue
+crop, use the full source image size, i.e. pixel array size.
+
+Horizontal and vertical blanking are specified by ``V4L2_CID_HBLANK`` and
+``V4L2_CID_VBLANK``, respectively. The unit of these controls are lines. The
+pixel rate is specified by ``V4L2_CID_PIXEL_RATE`` in the same sub-device. The
+unit of that control is Hz.
+
+Register list based drivers need to implement read-only sub-device nodes for the
+purpose. Devices that are not register list based need these to configure the
+device's internal processing pipeline.
+
+The first entity in the linear pipeline is the pixel array. The pixel array may
+be followed by other entities that are there to allow configuring binning,
+skipping, scaling or digital crop :ref:`v4l2-subdev-selections`.
+
+USB cameras etc. devices
+~~~~~~~~~~~~~~~~~~~~~~~~
+
+USB video class hardware, as well as many cameras offering a higher level
+control interface, generally use the concept of frame interval (or frame rate)
+on the level of device hardware interface. This means lower level controls
+exposed by raw cameras may not be used as an interface to control the frame
+interval on these devices.
+
+Power management
+----------------
+
+Always use runtime PM to manage the power states of your device. Camera sensor
+drivers are in no way special in this respect: they are responsible for
+controlling the power state of the device they otherwise control as well. In
+general, the device must be powered on at least when its registers are being
+accessed and when it is streaming.
+
+Existing camera sensor drivers may rely on the old
+:c:type:`v4l2_subdev_core_ops`->s_power() callback for bridge or ISP drivers to
+manage their power state. This is however **deprecated**. If you feel you need
+to begin calling an s_power from an ISP or a bridge driver, instead please add
+runtime PM support to the sensor driver you are using. Likewise, new drivers
+should not use s_power.
+
+Please see examples in e.g. ``drivers/media/i2c/ov8856.c`` and
+``drivers/media/i2c/smiapp/smiapp-core.c``. The two drivers work in both ACPI
+and DT based systems.
+
+Control framework
+~~~~~~~~~~~~~~~~~
+
+``v4l2_ctrl_handler_setup()`` function may not be used in the device's runtime
+PM ``runtime_resume`` callback, as it has no way to figure out the power state
+of the device. This is because the power state of the device is only changed
+after the power state transition has taken place. The ``s_ctrl``callback can be
+used to obtain device's power state after the power state transition:
+
+.. c:function::
+	int pm_runtime_get_if_in_use(struct device *dev);
+
+The function returns a non-zero value if it succeeded getting the power count or
+runtime PM was disabled, in either of which cases the driver may proceed to
+access the device.
diff --git a/Documentation/driver-api/media/csi2.rst b/Documentation/driver-api/media/csi2.rst
index 17cad435f1a0..e1b838014906 100644
--- a/Documentation/driver-api/media/csi2.rst
+++ b/Documentation/driver-api/media/csi2.rst
@@ -1,5 +1,7 @@
 .. SPDX-License-Identifier: GPL-2.0
 
+.. _MIPI_CSI_2:
+
 MIPI CSI-2
 ==========
 
diff --git a/Documentation/driver-api/media/index.rst b/Documentation/driver-api/media/index.rst
index 328350924853..c140692454b1 100644
--- a/Documentation/driver-api/media/index.rst
+++ b/Documentation/driver-api/media/index.rst
@@ -34,6 +34,7 @@ Please see:
     mc-core
     cec-core
     csi2
+    camera-sensor
 
     drivers/index
 
-- 
2.27.0


^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH v3 1/1] Documentation: media: Document how to write camera sensor drivers
  2020-07-30 16:20 [PATCH v3 1/1] Documentation: media: Document how to write camera sensor drivers Sakari Ailus
@ 2020-08-03 15:25 ` Helen Koike
  2020-08-03 21:25   ` Sakari Ailus
  0 siblings, 1 reply; 5+ messages in thread
From: Helen Koike @ 2020-08-03 15:25 UTC (permalink / raw)
  To: Sakari Ailus, linux-media; +Cc: laurent.pinchart, hverkuil

Hi Sakari,

Thanks for this.

On 7/30/20 1:20 PM, Sakari Ailus wrote:
> While we have had some example drivers, there has been up to date no
> formal documentation on how camera sensor drivers should be written; what
> are the practices, why, and where they apply.
> 
> Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
> ---
> since v2:
> 
> - More verbose explanation on sensor driver's responsibilities.
> 
> - Reword the explanation on power state vs. v4l2_ctrl_handler_setup.
> 
>  .../driver-api/media/camera-sensor.rst        | 134 ++++++++++++++++++
>  Documentation/driver-api/media/csi2.rst       |   2 +
>  Documentation/driver-api/media/index.rst      |   1 +
>  3 files changed, 137 insertions(+)
>  create mode 100644 Documentation/driver-api/media/camera-sensor.rst
> 
> diff --git a/Documentation/driver-api/media/camera-sensor.rst b/Documentation/driver-api/media/camera-sensor.rst
> new file mode 100644
> index 000000000000..2121586e8ede
> --- /dev/null
> +++ b/Documentation/driver-api/media/camera-sensor.rst
> @@ -0,0 +1,134 @@
> +.. SPDX-License-Identifier: GPL-2.0
> +
> +Writing camera sensor drivers
> +=============================
> +
> +CSI-2
> +-----
> +
> +Please see what is written on :ref:`MIPI_CSI_2`.
> +
> +Handling clocks
> +---------------
> +
> +Camera sensors have an internal clock tree including a PLL and a number of
> +divisors. The clock tree is generally configured by the driver based on a few
> +input parameters that are specific to the hardware:: the external clock frequency

Double colon

> +and the link frequency. The two parameters generally are obtained from system
> +firmware. No other frequencies should be used in any circumstances.

If I understand correctly, the firmware exposes a list of options that userspace can chose from, right?

> +
> +The reason why the clock frequencies are so important is that the clock signals
> +come out of the SoC, and in many cases a specific frequency is designed to be
> +used in the system. Using another frequency may cause harmful effects
> +elsewhere. Therefore only the pre-determined frequencies are configurable by the
> +user.
> +
> +Frame size
> +----------
> +
> +There are two distinct ways to configure the frame size produced by camera
> +sensors.
> +
> +Freely configurable camera sensor drivers
> +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> +
> +Freely configurable camera sensor drivers expose the device's internal
> +processing pipeline as one or more sub-devices with different cropping and
> +scaling configurations. The output size of the device is the result of a series
> +of cropping and scaling operations from the device's pixel array's size.
> +
> +An example of such a driver is the smiapp driver (see drivers/media/i2c/smiapp).
> +
> +Register list based drivers
> +~~~~~~~~~~~~~~~~~~~~~~~~~~~
> +
> +Register list based drivers generally, instead of able to configure the device
> +they control based on user requests, are limited to a number of preset
> +configurations that combine a number of different parameters that on hardware
> +level are independent. How a driver picks such configuration is based on the
> +format set on a source pad at the end of the device's internal pipeline.

If I understand correctly, the difference between the first and second way,
is that the first allows userspace to change the format on the subdevice's pads,
and the second one needs to be calculated from the format set in the video devices,
is this correct?

> +
> +Most sensor drivers are implemented this way, see e.g.
> +drivers/media/i2c/imx319.c for an example.
> +
> +Frame interval configuration
> +----------------------------
> +
> +There are two different methods for obtaining possibilities for different frame
> +intervals as well as configuring the frame interval. Which one to implement
> +depends on the type of the device.
> +
> +Raw camera sensors
> +~~~~~~~~~~~~~~~~~~
> +
> +Instead of a high level parameter such as frame interval, the frame interval is
> +a result of the configuration of a number of camera sensor implementation
> +specific parameters. Luckily, these parameters tend to be the same for more or
> +less all modern raw camera sensors.
> +
> +The frame interval is calculated using the following equation::
> +
> +	frame interval = (analogue crop width + horizontal blanking) *
> +			 (analogue crop height + vertical blanking) / pixel rate
> +
> +The formula is bus independent and is applicable for raw timing parameters on
> +large variety of devices beyond camera sensors. Devices that have no analogue
> +crop, use the full source image size, i.e. pixel array size.
> +
> +Horizontal and vertical blanking are specified by ``V4L2_CID_HBLANK`` and
> +``V4L2_CID_VBLANK``, respectively. The unit of these controls are lines. The
> +pixel rate is specified by ``V4L2_CID_PIXEL_RATE`` in the same sub-device. The
> +unit of that control is Hz.
> +
> +Register list based drivers need to implement read-only sub-device nodes for the
> +purpose. Devices that are not register list based need these to configure the
> +device's internal processing pipeline.
> +
> +The first entity in the linear pipeline is the pixel array. The pixel array may

A pixel array entity is the one that represents the "source" of the image, right?

> +be followed by other entities that are there to allow configuring binning,
> +skipping, scaling or digital crop :ref:`v4l2-subdev-selections`.
> +
> +USB cameras etc. devices
> +~~~~~~~~~~~~~~~~~~~~~~~~
> +
> +USB video class hardware, as well as many cameras offering a higher level
> +control interface, generally use the concept of frame interval (or frame rate)
> +on the level of device hardware interface. This means lower level controls
> +exposed by raw cameras may not be used as an interface to control the frame
> +interval on these devices.

Sorry, it's not clear to me if you are referring to driver->hardware configuration, or
userspace->driver configuration (but maybe it's just me and my lack of knowledge).

Regards,
Helen

> +
> +Power management
> +----------------
> +
> +Always use runtime PM to manage the power states of your device. Camera sensor
> +drivers are in no way special in this respect: they are responsible for
> +controlling the power state of the device they otherwise control as well. In
> +general, the device must be powered on at least when its registers are being
> +accessed and when it is streaming.
> +
> +Existing camera sensor drivers may rely on the old
> +:c:type:`v4l2_subdev_core_ops`->s_power() callback for bridge or ISP drivers to
> +manage their power state. This is however **deprecated**. If you feel you need
> +to begin calling an s_power from an ISP or a bridge driver, instead please add
> +runtime PM support to the sensor driver you are using. Likewise, new drivers
> +should not use s_power.
> +
> +Please see examples in e.g. ``drivers/media/i2c/ov8856.c`` and
> +``drivers/media/i2c/smiapp/smiapp-core.c``. The two drivers work in both ACPI
> +and DT based systems.
> +
> +Control framework
> +~~~~~~~~~~~~~~~~~
> +
> +``v4l2_ctrl_handler_setup()`` function may not be used in the device's runtime
> +PM ``runtime_resume`` callback, as it has no way to figure out the power state
> +of the device. This is because the power state of the device is only changed
> +after the power state transition has taken place. The ``s_ctrl``callback can be
> +used to obtain device's power state after the power state transition:
> +
> +.. c:function::
> +	int pm_runtime_get_if_in_use(struct device *dev);
> +
> +The function returns a non-zero value if it succeeded getting the power count or
> +runtime PM was disabled, in either of which cases the driver may proceed to
> +access the device.
> diff --git a/Documentation/driver-api/media/csi2.rst b/Documentation/driver-api/media/csi2.rst
> index 17cad435f1a0..e1b838014906 100644
> --- a/Documentation/driver-api/media/csi2.rst
> +++ b/Documentation/driver-api/media/csi2.rst
> @@ -1,5 +1,7 @@
>  .. SPDX-License-Identifier: GPL-2.0
>  
> +.. _MIPI_CSI_2:
> +
>  MIPI CSI-2
>  ==========
>  
> diff --git a/Documentation/driver-api/media/index.rst b/Documentation/driver-api/media/index.rst
> index 328350924853..c140692454b1 100644
> --- a/Documentation/driver-api/media/index.rst
> +++ b/Documentation/driver-api/media/index.rst
> @@ -34,6 +34,7 @@ Please see:
>      mc-core
>      cec-core
>      csi2
> +    camera-sensor
>  
>      drivers/index
>  
> 

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH v3 1/1] Documentation: media: Document how to write camera sensor drivers
  2020-08-03 15:25 ` Helen Koike
@ 2020-08-03 21:25   ` Sakari Ailus
  2020-08-03 21:39     ` Helen Koike
  0 siblings, 1 reply; 5+ messages in thread
From: Sakari Ailus @ 2020-08-03 21:25 UTC (permalink / raw)
  To: Helen Koike; +Cc: linux-media, laurent.pinchart, hverkuil

Hi Helen,

On Mon, Aug 03, 2020 at 12:25:54PM -0300, Helen Koike wrote:
> Hi Sakari,
> 
> Thanks for this.

Thanks for the comments!

> 
> On 7/30/20 1:20 PM, Sakari Ailus wrote:
> > While we have had some example drivers, there has been up to date no
> > formal documentation on how camera sensor drivers should be written; what
> > are the practices, why, and where they apply.
> > 
> > Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
> > ---
> > since v2:
> > 
> > - More verbose explanation on sensor driver's responsibilities.
> > 
> > - Reword the explanation on power state vs. v4l2_ctrl_handler_setup.
> > 
> >  .../driver-api/media/camera-sensor.rst        | 134 ++++++++++++++++++
> >  Documentation/driver-api/media/csi2.rst       |   2 +
> >  Documentation/driver-api/media/index.rst      |   1 +
> >  3 files changed, 137 insertions(+)
> >  create mode 100644 Documentation/driver-api/media/camera-sensor.rst
> > 
> > diff --git a/Documentation/driver-api/media/camera-sensor.rst b/Documentation/driver-api/media/camera-sensor.rst
> > new file mode 100644
> > index 000000000000..2121586e8ede
> > --- /dev/null
> > +++ b/Documentation/driver-api/media/camera-sensor.rst
> > @@ -0,0 +1,134 @@
> > +.. SPDX-License-Identifier: GPL-2.0
> > +
> > +Writing camera sensor drivers
> > +=============================
> > +
> > +CSI-2
> > +-----
> > +
> > +Please see what is written on :ref:`MIPI_CSI_2`.
> > +
> > +Handling clocks
> > +---------------
> > +
> > +Camera sensors have an internal clock tree including a PLL and a number of
> > +divisors. The clock tree is generally configured by the driver based on a few
> > +input parameters that are specific to the hardware:: the external clock frequency
> 
> Double colon

This is ReST. This should translate into a single colon.

> 
> > +and the link frequency. The two parameters generally are obtained from system
> > +firmware. No other frequencies should be used in any circumstances.
> 
> If I understand correctly, the firmware exposes a list of options that
> userspace can chose from, right?

Correct.

> 
> > +
> > +The reason why the clock frequencies are so important is that the clock signals
> > +come out of the SoC, and in many cases a specific frequency is designed to be
> > +used in the system. Using another frequency may cause harmful effects
> > +elsewhere. Therefore only the pre-determined frequencies are configurable by the
> > +user.
> > +
> > +Frame size
> > +----------
> > +
> > +There are two distinct ways to configure the frame size produced by camera
> > +sensors.
> > +
> > +Freely configurable camera sensor drivers
> > +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > +
> > +Freely configurable camera sensor drivers expose the device's internal
> > +processing pipeline as one or more sub-devices with different cropping and
> > +scaling configurations. The output size of the device is the result of a series
> > +of cropping and scaling operations from the device's pixel array's size.
> > +
> > +An example of such a driver is the smiapp driver (see drivers/media/i2c/smiapp).
> > +
> > +Register list based drivers
> > +~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > +
> > +Register list based drivers generally, instead of able to configure the device
> > +they control based on user requests, are limited to a number of preset
> > +configurations that combine a number of different parameters that on hardware
> > +level are independent. How a driver picks such configuration is based on the
> > +format set on a source pad at the end of the device's internal pipeline.
> 
> If I understand correctly, the difference between the first and second way,
> is that the first allows userspace to change the format on the subdevice's pads,
> and the second one needs to be calculated from the format set in the video devices,
> is this correct?

Yes, but video devices aren't necessarily involved. In a MC-centric drivers
the interface is available through sub-device nodes in both above cases.

> 
> > +
> > +Most sensor drivers are implemented this way, see e.g.
> > +drivers/media/i2c/imx319.c for an example.
> > +
> > +Frame interval configuration
> > +----------------------------
> > +
> > +There are two different methods for obtaining possibilities for different frame
> > +intervals as well as configuring the frame interval. Which one to implement
> > +depends on the type of the device.
> > +
> > +Raw camera sensors
> > +~~~~~~~~~~~~~~~~~~
> > +
> > +Instead of a high level parameter such as frame interval, the frame interval is
> > +a result of the configuration of a number of camera sensor implementation
> > +specific parameters. Luckily, these parameters tend to be the same for more or
> > +less all modern raw camera sensors.
> > +
> > +The frame interval is calculated using the following equation::
> > +
> > +	frame interval = (analogue crop width + horizontal blanking) *
> > +			 (analogue crop height + vertical blanking) / pixel rate
> > +
> > +The formula is bus independent and is applicable for raw timing parameters on
> > +large variety of devices beyond camera sensors. Devices that have no analogue
> > +crop, use the full source image size, i.e. pixel array size.
> > +
> > +Horizontal and vertical blanking are specified by ``V4L2_CID_HBLANK`` and
> > +``V4L2_CID_VBLANK``, respectively. The unit of these controls are lines. The
> > +pixel rate is specified by ``V4L2_CID_PIXEL_RATE`` in the same sub-device. The
> > +unit of that control is Hz.
> > +
> > +Register list based drivers need to implement read-only sub-device nodes for the
> > +purpose. Devices that are not register list based need these to configure the
> > +device's internal processing pipeline.
> > +
> > +The first entity in the linear pipeline is the pixel array. The pixel array may
> 
> A pixel array entity is the one that represents the "source" of the image, right?

Correct.

Please wrap lines before 80.

> 
> > +be followed by other entities that are there to allow configuring binning,
> > +skipping, scaling or digital crop :ref:`v4l2-subdev-selections`.
> > +
> > +USB cameras etc. devices
> > +~~~~~~~~~~~~~~~~~~~~~~~~
> > +
> > +USB video class hardware, as well as many cameras offering a higher level
> > +control interface, generally use the concept of frame interval (or frame rate)
> > +on the level of device hardware interface. This means lower level controls
> > +exposed by raw cameras may not be used as an interface to control the frame
> > +interval on these devices.
> 
> Sorry, it's not clear to me if you are referring to driver->hardware configuration, or
> userspace->driver configuration (but maybe it's just me and my lack of knowledge).

Both actually. Would this be clearer:

USB video class hardware, as well as many cameras offering a similar higher
level interface natively, generally use the concept of frame interval (or
frame rate) on device level in firmware or hardware. This means lower level
controls implemented by raw cameras may not be used on uAPI (or even kAPI)
to control the frame interval on these devices.

-- 
Kind regards,

Sakari Ailus

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH v3 1/1] Documentation: media: Document how to write camera sensor drivers
  2020-08-03 21:25   ` Sakari Ailus
@ 2020-08-03 21:39     ` Helen Koike
  2020-08-03 22:49       ` Sakari Ailus
  0 siblings, 1 reply; 5+ messages in thread
From: Helen Koike @ 2020-08-03 21:39 UTC (permalink / raw)
  To: Sakari Ailus; +Cc: linux-media, laurent.pinchart, hverkuil



On 8/3/20 6:25 PM, Sakari Ailus wrote:
> Hi Helen,
> 
> On Mon, Aug 03, 2020 at 12:25:54PM -0300, Helen Koike wrote:
>> Hi Sakari,
>>
>> Thanks for this.
> 
> Thanks for the comments!
> 
>>
>> On 7/30/20 1:20 PM, Sakari Ailus wrote:
>>> While we have had some example drivers, there has been up to date no
>>> formal documentation on how camera sensor drivers should be written; what
>>> are the practices, why, and where they apply.
>>>
>>> Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
>>> ---
>>> since v2:
>>>
>>> - More verbose explanation on sensor driver's responsibilities.
>>>
>>> - Reword the explanation on power state vs. v4l2_ctrl_handler_setup.
>>>
>>>  .../driver-api/media/camera-sensor.rst        | 134 ++++++++++++++++++
>>>  Documentation/driver-api/media/csi2.rst       |   2 +
>>>  Documentation/driver-api/media/index.rst      |   1 +
>>>  3 files changed, 137 insertions(+)
>>>  create mode 100644 Documentation/driver-api/media/camera-sensor.rst
>>>
>>> diff --git a/Documentation/driver-api/media/camera-sensor.rst b/Documentation/driver-api/media/camera-sensor.rst
>>> new file mode 100644
>>> index 000000000000..2121586e8ede
>>> --- /dev/null
>>> +++ b/Documentation/driver-api/media/camera-sensor.rst
>>> @@ -0,0 +1,134 @@
>>> +.. SPDX-License-Identifier: GPL-2.0
>>> +
>>> +Writing camera sensor drivers
>>> +=============================
>>> +
>>> +CSI-2
>>> +-----
>>> +
>>> +Please see what is written on :ref:`MIPI_CSI_2`.
>>> +
>>> +Handling clocks
>>> +---------------
>>> +
>>> +Camera sensors have an internal clock tree including a PLL and a number of
>>> +divisors. The clock tree is generally configured by the driver based on a few
>>> +input parameters that are specific to the hardware:: the external clock frequency
>>
>> Double colon
> 
> This is ReST. This should translate into a single colon.
> 
>>
>>> +and the link frequency. The two parameters generally are obtained from system
>>> +firmware. No other frequencies should be used in any circumstances.
>>
>> If I understand correctly, the firmware exposes a list of options that
>> userspace can chose from, right?
> 
> Correct.
> 
>>
>>> +
>>> +The reason why the clock frequencies are so important is that the clock signals
>>> +come out of the SoC, and in many cases a specific frequency is designed to be
>>> +used in the system. Using another frequency may cause harmful effects
>>> +elsewhere. Therefore only the pre-determined frequencies are configurable by the
>>> +user.
>>> +
>>> +Frame size
>>> +----------
>>> +
>>> +There are two distinct ways to configure the frame size produced by camera
>>> +sensors.
>>> +
>>> +Freely configurable camera sensor drivers
>>> +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>> +
>>> +Freely configurable camera sensor drivers expose the device's internal
>>> +processing pipeline as one or more sub-devices with different cropping and
>>> +scaling configurations. The output size of the device is the result of a series
>>> +of cropping and scaling operations from the device's pixel array's size.
>>> +
>>> +An example of such a driver is the smiapp driver (see drivers/media/i2c/smiapp).
>>> +
>>> +Register list based drivers
>>> +~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>> +
>>> +Register list based drivers generally, instead of able to configure the device
>>> +they control based on user requests, are limited to a number of preset
>>> +configurations that combine a number of different parameters that on hardware
>>> +level are independent. How a driver picks such configuration is based on the
>>> +format set on a source pad at the end of the device's internal pipeline.
>>
>> If I understand correctly, the difference between the first and second way,
>> is that the first allows userspace to change the format on the subdevice's pads,
>> and the second one needs to be calculated from the format set in the video devices,
>> is this correct?
> 
> Yes, but video devices aren't necessarily involved. In a MC-centric drivers
> the interface is available through sub-device nodes in both above cases.

So the definition of a "Freely configurable camera sensor driver" is when
we have MC-centric driver exposing format configuration through pads, and
a "Register list based driver" is when the format configuration is propagated
in the pipeline from somewhere else right? And it can be a MC-centric device
where it doesn't expose format configuration through pads, or non Mc-centric,
where the subdevice is not exposed, is this correct?

I was wondering why it is called "Register list".

> 
>>
>>> +
>>> +Most sensor drivers are implemented this way, see e.g.
>>> +drivers/media/i2c/imx319.c for an example.
>>> +
>>> +Frame interval configuration
>>> +----------------------------
>>> +
>>> +There are two different methods for obtaining possibilities for different frame
>>> +intervals as well as configuring the frame interval. Which one to implement
>>> +depends on the type of the device.
>>> +
>>> +Raw camera sensors
>>> +~~~~~~~~~~~~~~~~~~
>>> +
>>> +Instead of a high level parameter such as frame interval, the frame interval is
>>> +a result of the configuration of a number of camera sensor implementation
>>> +specific parameters. Luckily, these parameters tend to be the same for more or
>>> +less all modern raw camera sensors.
>>> +
>>> +The frame interval is calculated using the following equation::
>>> +
>>> +	frame interval = (analogue crop width + horizontal blanking) *
>>> +			 (analogue crop height + vertical blanking) / pixel rate
>>> +
>>> +The formula is bus independent and is applicable for raw timing parameters on
>>> +large variety of devices beyond camera sensors. Devices that have no analogue
>>> +crop, use the full source image size, i.e. pixel array size.
>>> +
>>> +Horizontal and vertical blanking are specified by ``V4L2_CID_HBLANK`` and
>>> +``V4L2_CID_VBLANK``, respectively. The unit of these controls are lines. The
>>> +pixel rate is specified by ``V4L2_CID_PIXEL_RATE`` in the same sub-device. The
>>> +unit of that control is Hz.
>>> +
>>> +Register list based drivers need to implement read-only sub-device nodes for the
>>> +purpose. Devices that are not register list based need these to configure the
>>> +device's internal processing pipeline.
>>> +
>>> +The first entity in the linear pipeline is the pixel array. The pixel array may
>>
>> A pixel array entity is the one that represents the "source" of the image, right?
> 
> Correct.
> 
> Please wrap lines before 80.
> 
>>
>>> +be followed by other entities that are there to allow configuring binning,
>>> +skipping, scaling or digital crop :ref:`v4l2-subdev-selections`.
>>> +
>>> +USB cameras etc. devices
>>> +~~~~~~~~~~~~~~~~~~~~~~~~
>>> +
>>> +USB video class hardware, as well as many cameras offering a higher level
>>> +control interface, generally use the concept of frame interval (or frame rate)
>>> +on the level of device hardware interface. This means lower level controls
>>> +exposed by raw cameras may not be used as an interface to control the frame
>>> +interval on these devices.
>>
>> Sorry, it's not clear to me if you are referring to driver->hardware configuration, or
>> userspace->driver configuration (but maybe it's just me and my lack of knowledge).
> 
> Both actually. Would this be clearer:
> 
> USB video class hardware, as well as many cameras offering a similar higher
> level interface natively, generally use the concept of frame interval (or
> frame rate) on device level in firmware or hardware. This means lower level
> controls implemented by raw cameras may not be used on uAPI (or even kAPI)
> to control the frame interval on these devices.
> 

Yes :)

Thanks,
Helen

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH v3 1/1] Documentation: media: Document how to write camera sensor drivers
  2020-08-03 21:39     ` Helen Koike
@ 2020-08-03 22:49       ` Sakari Ailus
  0 siblings, 0 replies; 5+ messages in thread
From: Sakari Ailus @ 2020-08-03 22:49 UTC (permalink / raw)
  To: Helen Koike; +Cc: linux-media, laurent.pinchart, hverkuil

Hi Helen,

On Mon, Aug 03, 2020 at 06:39:23PM -0300, Helen Koike wrote:
> 
> 
> On 8/3/20 6:25 PM, Sakari Ailus wrote:
> > Hi Helen,
> > 
> > On Mon, Aug 03, 2020 at 12:25:54PM -0300, Helen Koike wrote:
> >> Hi Sakari,
> >>
> >> Thanks for this.
> > 
> > Thanks for the comments!
> > 
> >>
> >> On 7/30/20 1:20 PM, Sakari Ailus wrote:
> >>> While we have had some example drivers, there has been up to date no
> >>> formal documentation on how camera sensor drivers should be written; what
> >>> are the practices, why, and where they apply.
> >>>
> >>> Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
> >>> ---
> >>> since v2:
> >>>
> >>> - More verbose explanation on sensor driver's responsibilities.
> >>>
> >>> - Reword the explanation on power state vs. v4l2_ctrl_handler_setup.
> >>>
> >>>  .../driver-api/media/camera-sensor.rst        | 134 ++++++++++++++++++
> >>>  Documentation/driver-api/media/csi2.rst       |   2 +
> >>>  Documentation/driver-api/media/index.rst      |   1 +
> >>>  3 files changed, 137 insertions(+)
> >>>  create mode 100644 Documentation/driver-api/media/camera-sensor.rst
> >>>
> >>> diff --git a/Documentation/driver-api/media/camera-sensor.rst b/Documentation/driver-api/media/camera-sensor.rst
> >>> new file mode 100644
> >>> index 000000000000..2121586e8ede
> >>> --- /dev/null
> >>> +++ b/Documentation/driver-api/media/camera-sensor.rst
> >>> @@ -0,0 +1,134 @@
> >>> +.. SPDX-License-Identifier: GPL-2.0
> >>> +
> >>> +Writing camera sensor drivers
> >>> +=============================
> >>> +
> >>> +CSI-2
> >>> +-----
> >>> +
> >>> +Please see what is written on :ref:`MIPI_CSI_2`.
> >>> +
> >>> +Handling clocks
> >>> +---------------
> >>> +
> >>> +Camera sensors have an internal clock tree including a PLL and a number of
> >>> +divisors. The clock tree is generally configured by the driver based on a few
> >>> +input parameters that are specific to the hardware:: the external clock frequency
> >>
> >> Double colon
> > 
> > This is ReST. This should translate into a single colon.
> > 
> >>
> >>> +and the link frequency. The two parameters generally are obtained from system
> >>> +firmware. No other frequencies should be used in any circumstances.
> >>
> >> If I understand correctly, the firmware exposes a list of options that
> >> userspace can chose from, right?
> > 
> > Correct.
> > 
> >>
> >>> +
> >>> +The reason why the clock frequencies are so important is that the clock signals
> >>> +come out of the SoC, and in many cases a specific frequency is designed to be
> >>> +used in the system. Using another frequency may cause harmful effects
> >>> +elsewhere. Therefore only the pre-determined frequencies are configurable by the
> >>> +user.
> >>> +
> >>> +Frame size
> >>> +----------
> >>> +
> >>> +There are two distinct ways to configure the frame size produced by camera
> >>> +sensors.
> >>> +
> >>> +Freely configurable camera sensor drivers
> >>> +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> >>> +
> >>> +Freely configurable camera sensor drivers expose the device's internal
> >>> +processing pipeline as one or more sub-devices with different cropping and
> >>> +scaling configurations. The output size of the device is the result of a series
> >>> +of cropping and scaling operations from the device's pixel array's size.
> >>> +
> >>> +An example of such a driver is the smiapp driver (see drivers/media/i2c/smiapp).
> >>> +
> >>> +Register list based drivers
> >>> +~~~~~~~~~~~~~~~~~~~~~~~~~~~
> >>> +
> >>> +Register list based drivers generally, instead of able to configure the device
> >>> +they control based on user requests, are limited to a number of preset
> >>> +configurations that combine a number of different parameters that on hardware
> >>> +level are independent. How a driver picks such configuration is based on the
> >>> +format set on a source pad at the end of the device's internal pipeline.
> >>
> >> If I understand correctly, the difference between the first and second way,
> >> is that the first allows userspace to change the format on the subdevice's pads,
> >> and the second one needs to be calculated from the format set in the video devices,
> >> is this correct?
> > 
> > Yes, but video devices aren't necessarily involved. In a MC-centric drivers
> > the interface is available through sub-device nodes in both above cases.
> 
> So the definition of a "Freely configurable camera sensor driver" is when
> we have MC-centric driver exposing format configuration through pads, and
> a "Register list based driver" is when the format configuration is propagated

On a register list based driver it's all configured on a single pad.

The sensor internal pipeline exists in both cases but for the latter class
of drivers it's not freely configurable.

Freely configurable drivers require an MC centric bridge driver to expose
the sensor internal pipeline as multiple sub-devices.

> in the pipeline from somewhere else right? And it can be a MC-centric device
> where it doesn't expose format configuration through pads, or non Mc-centric,
> where the subdevice is not exposed, is this correct?
> 
> I was wondering why it is called "Register list".

These drivers more or less have large lists of register address--value
pairs the driver writes to the device at a suitable moment. Such drivers
know very little about the device they control.

...

> >>> +be followed by other entities that are there to allow configuring binning,
> >>> +skipping, scaling or digital crop :ref:`v4l2-subdev-selections`.
> >>> +
> >>> +USB cameras etc. devices
> >>> +~~~~~~~~~~~~~~~~~~~~~~~~
> >>> +
> >>> +USB video class hardware, as well as many cameras offering a higher level
> >>> +control interface, generally use the concept of frame interval (or frame rate)
> >>> +on the level of device hardware interface. This means lower level controls
> >>> +exposed by raw cameras may not be used as an interface to control the frame
> >>> +interval on these devices.
> >>
> >> Sorry, it's not clear to me if you are referring to driver->hardware configuration, or
> >> userspace->driver configuration (but maybe it's just me and my lack of knowledge).
> > 
> > Both actually. Would this be clearer:
> > 
> > USB video class hardware, as well as many cameras offering a similar higher
> > level interface natively, generally use the concept of frame interval (or
> > frame rate) on device level in firmware or hardware. This means lower level
> > controls implemented by raw cameras may not be used on uAPI (or even kAPI)
> > to control the frame interval on these devices.
> > 
> 
> Yes :)

Thanks. I'll use this in v4.

-- 
Kind regards,

Sakari Ailus

^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, back to index

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-07-30 16:20 [PATCH v3 1/1] Documentation: media: Document how to write camera sensor drivers Sakari Ailus
2020-08-03 15:25 ` Helen Koike
2020-08-03 21:25   ` Sakari Ailus
2020-08-03 21:39     ` Helen Koike
2020-08-03 22:49       ` Sakari Ailus

Linux-Media Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-media/0 linux-media/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-media linux-media/ https://lore.kernel.org/linux-media \
		linux-media@vger.kernel.org
	public-inbox-index linux-media

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-media


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git